continuous-learning

Extracting reusable patterns from Claude Code sessions involves evaluating end-of-session transcripts to identify repeatable issues and approaches. The skill stores these extracted patterns as learned skills in ~/.claude/skills/learned/ for future workflows.

Automatic pattern extraction from session transcripts is a technique that analyzes end-of-session logs to identify repeatable fixes and approaches. It transforms overlooked session data into stored, learned skills for future Claude workflows.

Yes, you need jq installed as a dependency to extract learned skills from Claude workflows. The skill relies on jq scripts to parse session transcripts and save the identified patterns to the learned skills directory.

Yes, you can configure the extraction threshold for Claude Code session evaluation. The skill supports configurable min_session_length, extraction_threshold, auto_approve, and learned_skills_path settings to control pattern extraction behavior.

The best way to automate learning from Claude Code sessions is integrating a Stop hook that triggers evaluation at session end. This automatically extracts repeatable patterns with minimal latency and saves them as learned skills.

Extracted patterns are stored as learned skills in the ~/.claude/skills/learned/ directory after analyzing Claude Code transcripts. This allows the identified repeatable approaches to be reused in future Claude workflows.